insider trading preceding goodwill impairments › faculty › publication files › 10-007.pdf ·...
TRANSCRIPT
Copyright © 2009 by Karl A. Muller, III, Monica Neamtiu, and Edward J. Riedl
Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may not be reproduced without permission of the copyright holder. Copies of working papers are available from the author.
Insider Trading Preceding Goodwill Impairments Karl A. Muller, III Monica Neamtiu Edward J. Riedl
Working Paper
10-007
Insider Trading Preceding Goodwill Impairments
Karl A. Muller, III
The Pennsylvania State University Smeal College of Business
384 Business Building University Park, PA 16802
Monica Neamtiu
University of Arizona Eller College of Management
301 McClelland Hall Tucson, AZ 85721
Edward J. Riedl
Harvard Business School 365 Morgan Hall
Boston, MA 02163 [email protected]
ABSTRACT: We investigate whether insiders strategically sell shares prior to the disclosure of goodwill impairment losses. We provide evidence that insiders of goodwill impairment firms engage in abnormal selling of their shares quarters prior to the announcement of such losses. In addition, of firms recording goodwill impairments, we provide evidence that those firms with insiders selling prior to the announcement of the loss face significantly more negative abnormal returns. Our findings are robust to subsample analysis examining firms reporting goodwill impairments and having low quality information environments (i.e., delayed price discovery). This isolates a setting wherein observed strategic trading behavior more likely reflects insiders’ private information regarding goodwill, as opposed to other (non-goodwill related) economic performance. Overall, the results are consistent with corporate insiders being able to profit from their private information relating to a specific financial reporting element, goodwill impairments, prior to its incorporation by the equity market or recognition by the firm’s accounting system.
Keywords: Goodwill, impairment, insider trading, SFAS 142
Current Date: July 2009
We thank Daniel Beneish, Francois Brochet, Jennifer Francis, Theodore Goodman, Trevor Harris, Leslie Hodder, Pat Hopkins, Amy Hutton, Michael Kimbrough, Ben Lansford, Peter Pope, Karthik Ramanna, Jim Vincent, Beverly Walther, and Joe Weber for valuable comments and discussions. We also appreciate useful comments from seminar participants at Indiana University, Lancaster University, Northwestern University, and the IMO Conference at Harvard Business School. Muller acknowledges financial support from the Smeal Faculty Fellowship for 2008-2009.
Insider Trading Preceding Goodwill Impairments
ABSTRACT: We investigate whether insiders strategically sell shares prior to the disclosure of goodwill impairment losses. We provide evidence that insiders of goodwill impairment firms engage in abnormal selling of their shares quarters prior to the announcement of such losses. In addition, of firms recording goodwill impairments, we provide evidence that those firms with insiders selling prior to the announcement of the loss face significantly more negative abnormal returns. Our findings are robust to subsample analysis examining firms reporting goodwill impairments and having low quality information environments (i.e., delayed price discovery). This isolates a setting wherein observed strategic trading behavior more likely reflects insiders’ private information regarding goodwill, as opposed to other (non-goodwill related) economic performance. Overall, the results are consistent with corporate insiders being able to profit from their private information relating to a specific financial reporting element, goodwill impairments, prior to its incorporation by the equity market or recognition by the firm’s accounting system.
Insider Trading Preceding Goodwill Impairments
I. INTRODUCTION
This study investigates whether firms’ insiders strategically sell their stock holdings prior
to the disclosure of goodwill impairment losses. Specifically, we examine whether insiders
engage in abnormal selling behavior in the quarters prior to the impairment disclosure, and
whether abnormal returns for firms with insiders engaging in selling behavior are relatively more
negative than the returns of other impairment firms whose insiders do not sell prior to the
disclosure. A number of recent studies provide evidence of insiders trading prior to the
announcement of earnings performance measures: future earnings surprises (Piotroski and
Roulstone 2005; Roulstone 2006; Henderson, Jagolinzer, and Muller 2009; Jagolinzer 2009), a
break in a string of consecutive quarterly earnings increases (Ke, Huddart, and Petroni 2003),
and price-relevant information disclosed in Form 10-K or 10-Q filings released following
earnings announcements (Huddart, Ke, and Shi 2007).1 What remains a puzzle, however, is the
types of information aggregated into reported earnings that constitute the source of insiders’
private information. Our study provides insights into this puzzle by investigating whether
insiders trade ahead of a specific component of reported earnings—goodwill impairments.
The setting of goodwill impairments provides a unique opportunity to investigate whether
firm insiders trade strategically prior to the release of a specific component of reported
accounting earnings. First, goodwill impairment charges tend to be economically important. In
our sample investigating goodwill impairments reported during 2002-2007 under Statement of
Financial Accounting Standards (SFAS) 142, Goodwill and Other Intangible Assets, such
charges average 11.9% of the market value of equity. Second, insiders likely possess private
1 Prior research also examines other aspects of the relation between insider trading and accounting information (e.g., Beneish and Vargus 2002; Jagolinzer and Roulstone 2008; Jayaraman 2008; and Rogers 2008).
2
information regarding the value of acquired goodwill. Managers frequently prepare internal
budgets for each reporting unit, which include information (e.g., earnings projections) that
directly relate to temporary and permanent declines in the value of acquired goodwill.2 Given
the proprietary nature of this information, however, such information is rarely publicly disclosed.
Finally, insiders’ private information advantage regarding the value of goodwill could be
relatively long-lived due to goodwill impairment testing rules delaying the accounting
recognition of economic goodwill impairments. Specifically, under SFAS 142, goodwill is tested
for impairment only when the entire reporting unit is impaired—which can provide a substantial
cushion if the fair value of the other assets in the reporting unit have risen above their carrying
value. In addition, goodwill impairment testing provides opportunities for managers to use their
discretion to further increase the cushion (e.g., through overestimation of future earnings) for
both the reporting units and for specific assets and liabilities.
Despite the potential for a significant private informational advantage by insiders, it is an
open question whether firm insiders trade on their private foreknowledge of goodwill
impairments. Significant strategic selling by insiders before goodwill write-offs may not occur
for two reasons. First, market prices may incorporate information regarding goodwill
impairments on a timely basis. Prior research (e.g., Gu and Lev 2008) argues that part of the
goodwill recorded at the time of purchase reflects overpayment in ill-advised acquisitions, and
that the characteristics of the original acquisitions can be powerful predictors of the eventual
goodwill write-offs. To the extent that market participants price any overpayment at the time of
acquisition, this can limit or eliminate how much private information managers possess 2 Under SFAS 142, testing for goodwill impairment is a two-step process performed at the reporting unit level. First, a firm assesses if a reporting unit’s fair value is less than its book value. If so, the firm then derives the implied fair value of goodwill by subtracting the fair value of the net assets of the reporting unit from the fair value of the reporting unit to obtain an implied fair value of the goodwill. If this implied fair value of goodwill is less than its book value, the difference is the goodwill impairment loss. In this process, internal budgets and earnings projections are critical in estimating the reporting unit and goodwill fair values.
3
regarding impaired goodwill. In addition, stock prices may incorporate information regarding
goodwill impairments relatively quickly as investors have access to firm and industry
performance measures, such as earnings and sales figures, which include prior reported and
future forecasted amounts. Second, even if insiders have incentives to sell before the
announcement of goodwill write-offs, legal and governance constraints and litigation concerns
may limit their ability to do so (e.g., Huddart, Ke, and Shi 2007).
We examine the trading of insiders of firms reporting goodwill impairment charges under
SFAS 142 during 2002-2007.3 Similar to prior insider trading studies investigating other
accounting events, our intent is to examine insiders’ trading ahead of the price discovery by the
market of the goodwill impairment, not just trading ahead of the short-window reaction to the
formal release of the accounting goodwill impairment. That is, we end our analyses on the
accounting release of the goodwill impairment, as little private information regarding the
impairment should exist beyond that point, but allow that the market may obtain information
regarding such losses months prior to the eventual release. We conduct our empirical analyses
using a sample of all goodwill impairments. In addition, we examine two subsamples split by
the quality of the information environment: low quality (defined as firms with below median
analyst following) and high quality (defined as firms with above median analyst following). Our
primary interest is in the low quality information environment, to isolate a setting in which price
discovery regarding goodwill impairments more likely occurs closer to its formal public release
(i.e., to mitigate the likelihood that trading on non-goodwill information explains our findings).
Consistent with prior research suggesting a positive relationship between analyst following and
the speed of price discovery (Brennan, Jegadeesh, and Swaminathan 1993; Brennan and
3 Our analyses focus on recognized goodwill impairments for a simple reason: they are observable, allowing empirical implementation. Similar strategic trading incentives exist for insiders of firms with impaired goodwill that avoid accounting loss recognition.
4
Subrahmanyam 1995; Kimbrough 2007), we expect in the low quality information environment
that lower analyst following results in relatively less competition amongst analysts and less
timely incorporation of information into market prices.4
Consistent with goodwill impairment charges leading to economically significant
negative returns, for goodwill impairment firms in the low quality information environment we
find that stock prices fall –5.53% (–2.44%) for the trading window of day –30 through +2 (day –
2 through + 2), where day 0 is the earnings announcement date of the goodwill impairment
quarter. In contrast, for goodwill impairment firms in the high quality information environment,
we find that stock prices fall –2.96% (–0.53%) for similarly measured windows. In addition,
consistent with goodwill impairment charges being impounded in prices later in a low quality
information environment, we find that over 56% of the total stock price decline over the 18
months prior to the goodwill impairment occurs during the trading window of day –30 through
+2 for firms in the low quality information environment. In contrast, we find that less than 12%
of the total stock price decline occurs over the same period for firms in the high quality
information environment.5
Regarding insiders’ net selling behavior, we find evidence that insiders of firms reporting
goodwill impairments are more likely relative to insiders of a group of control firms not
reporting goodwill impairments to sell their holdings up to twenty-four months prior to the
goodwill impairment announcement. Moreover, of firms recording goodwill impairments, firms
where insiders are net sellers exhibit significantly more negative abnormal returns than firms
where insiders are net buyers or do not trade shares. In addition, our results are robust to
subsample analysis focusing on firms having low quality information environments, consistent
4 A number of other studies find similar evidence of lower information asymmetries in markets when analyst following is higher (e.g., Hong, Lim, and Stein 2000; Barth and Hutton 2004; and Ellul and Panayides 2008). 5 Alternative windows (described later) provide similar inferences.
5
with the relatively less timely price discovery enabling insiders to better exploit any private
information. These findings are consistent with insiders possessing private information
regarding goodwill impairments that becomes incorporated into market prices in the quarters
following insiders’ sales.
These findings extend prior insider trading research investigating strategic trade prior to
the release of accounting information. Specifically, our study provides evidence of a particular
financial reporting item—goodwill impairments—that insiders appear to anticipate and trade on
far in advance of its public disclosure. In addition, our findings extend prior research
investigating managers’ discretion over goodwill impairments by documenting evidence of an
incentive for the delayed release of goodwill impairment amounts. Finally, our findings increase
our understanding of how the equity market learns about information pertaining to goodwill
impairments. Despite the relatively large magnitude of reported impairment losses, prior
literature finds insignificant or relatively small negative market reactions to the announcement of
these impairments (Francis, Hanna, and Vincent 1996; Bens, Heltzer, and Segal 2007). We
provide evidence that market prices reflect goodwill impairment losses more rapidly in high
quality information environments, and that market reactions to the announcement of goodwill
impairments are substantially higher in low quality information environments.
In the next section, we present background information and develop our hypotheses.
Section III includes our sample selection procedure and descriptive information. Section IV
presents the research design and the empirical findings. Section V concludes.
II. BACKGROUND AND RESEARCH HYPOTHESES
In this section, we discuss the background on reporting goodwill impairment losses. We
6
also discuss our expectations about the trading behavior of corporate insiders in the quarters
preceding goodwill impairment announcements. Regarding the latter, under the argument that
goodwill impairments are sources of non-public information, we discuss the incentives of
corporate insiders to sell their holdings prior to goodwill impairment announcements.
Background on Reporting Goodwill Impairment Losses
Accounting for goodwill has been the subject of debate for a number of years, reflected in
an evolving framework to address the reporting of this asset. The issuance of Accounting
Principles Board Opinion No. 17, Intangible Assets, in 1970 provided early guidance on the
reporting of goodwill, with two major provisions. First, goodwill was to be amortized over a
period not exceeding forty years; second, any impairment testing of goodwill would be
performed at the enterprise level. The subsequent issuance of SFAS 121, Accounting for the
Impairment of Long-Lived Assets and for Long-Lived Assets to Be Disposed Of, in 1995 sought
to provide more guidance of impairment testing. Specifically, goodwill impairments were only
assessed when an asset or a group of assets acquired in a business combination were suspected of
being impaired due to some “triggering” event (e.g., operating losses or change in intended use
of the asset). During periods when one or more triggering events were present at the lowest
cash-generating unit level, an impairment loss would occur if the recoverable value of the
assets—defined as the undiscounted future cash flows—exceeded the carrying value of the
assets. Most recently, SFAS 142, Goodwill and Other Intangible Assets, was issued in 2001 to
revisit the reporting of goodwill, with three salient provisions. First, goodwill is no longer
amortized, but must be tested annually for impairment. Second, the testing for impairment must
occur at the reporting unit level, which often aligns with a firm’s operating segment level
reported under the provisions of SFAS 131, Disclosures about Segments of an Enterprise and
7
Related Information. Finally, the impairment assessment is dictated under a two-step impairment
test as follows: assess if the reporting unit fair value is less than its book value; if so, then derive
the implied fair value of goodwill by subtracting the fair value of the net assets of the reporting
unit from the fair value of the reporting unit; if the implied fair value of the goodwill is less than
its book value, the difference is the reported goodwill impairment.
Research Hypotheses
Prior research provides evidence of corporate insiders strategically trading on news prior
to its public announcement. For instance, a relatively large number of studies provide evidence
of managers trading prior to the announcement of corporate events such as acquisitions (Seyhun
1990), bankruptcy (Seyhun and Bradley 1997), initiations of dividends (John and Lang 1991),
seasoned equity offerings (Karpoff and Lee 1991), and share repurchases (Lee, Mikkelson, and
Partch 1992). In addition, more recent studies provide evidence of insiders trading prior to the
announcement of earnings performance measures such as earnings announcements (Piotroski and
Roulstone 2005; Roulstone 2006; Henderson, Jagolinzer, and Muller 2009; Jagolinzer 2009), a
break in a string of consecutive quarterly earnings increases (Ke, Huddart, and Petroni 2003),
and price-relevant information disclosed in Form 10-K or 10-Q filings released following
earnings announcements (Huddart, Ke, and Shi 2007).6
What remains unexplored is the specific types of information aggregated in reported
earnings that are the source of managers’ private information. Goodwill impairment charges
represent a particularly promising earnings component that insiders could strategically trade
ahead of (i.e., prior to its incorporation by the equity market or its accounting recognition).
6 Earlier research (e.g., Elliot, Morse, and Richardson 1984; Givoly and Palmon 1985; Sivakumar and Waymire 1994) fails to find evidence of strategic trade by corporate insiders prior to earnings announcements.
8
Specifically, goodwill impairment losses typically represent sizable charges against earnings. In
addition, insiders have access to internal budget estimates, which are updated with regular
frequency. These budgets contain profitability forecasts that are critical in assessing the value of
recorded goodwill amounts, and are not typically disclosed externally. Finally, impairments in
the value of recorded goodwill could be a source of private information for insiders over
relatively extended time periods. Under SFAS 142, impairments of specific assets, such as
goodwill, are only recognized when the entire reporting unit is impaired―i.e., when the fair
value of the reporting unit falls below its carrying value. Thus, a considerable cushion regarding
the recording of goodwill impairment losses can exist if the fair value of the other assets in the
unit exceeds their carrying value. In addition, given that goodwill impairment testing requires
estimating fair values for both the reporting units and for specific assets and liabilities in the
absence of verifiable market prices, considerable managerial discretion can contribute to an even
greater cushion. The views of the practitioner literature express concerns over the managerial
discretion afforded by SFAS 142 (e.g., Hlousek 2002).
Consistent with managers having considerable discretion over the timing and
measurement of goodwill impairment losses under SFAS 142 and earlier financial reporting
standards, prior and concurrent academic research has found evidence of managers acting
opportunistically when recognizing goodwill impairment losses. In the first analysis of such
discretion, Francis, Hanna, and Vincent (1996) provides evidence that goodwill impairment
losses announced during 1989-1992 (a time-period lacking explicit guidance regarding goodwill
impairments) are substantially affected by reporting incentives such as those to manage earnings
upward or smooth earnings and, in the case of newly hired CEOs, to take a “big bath.” In
addition, the study fails to find a market reaction to announcements of goodwill impairment
9
losses, which is interpreted as the measurement of goodwill impairment losses being adversely
affected by managerial reporting incentives. More recently, studies investigating SFAS 121 and
142 goodwill impairment losses provide evidence of significant managerial influence over the
reporting of goodwill impairment losses (Riedl 2004; Beatty and Weber 2006; Anantharaman
2008; Ramanna and Watts 2008), but find some evidence of negative market reactions to the
announcement of goodwill impairment losses (Li, Schroff, and Venkataraman 2004; Ahmed and
Guler 2007; Bens, Heltzer, and Segal 2007).
Given that the goodwill impairment losses under SFAS 142 may be privately known by
corporate insiders far in advance of their market and accounting recognition, corporate insiders
face incentives to behave strategically with respect to their trades prior to the public
announcement of such losses. Assuming that corporate insiders privately infer that a goodwill
impairment loss has occurred and may ultimately be recognized in the financial reports in the
future, insiders will delay planned purchases (e.g., to achieve some targeted stock ownership
level) and accelerate planned sales (e.g., for personal liquidity needs). If the price impact of their
private information ends with public disclosure, these trading incentives will continue until the
goodwill impairment loss is recognized.
Despite these incentives for strategic trade by insiders, it is unclear ex ante that insiders
would unwind their holdings before the public announcement of a goodwill write-off. Corporate
insiders may not trade strategically in advance of goodwill impairment announcements for a
number of reasons. First, managers and other insiders may have little private information
regarding goodwill impairment losses prior to their recognition. For instance, investors may
infer goodwill impairment losses from alternative sources of information. One possible source is
the characteristics of the original acquisition that gave rise to the goodwill. Gu and Lev (2008)
10
provides evidence that factors suggesting overpayment in acquisitions can be powerful predictors
of future goodwill impairments. Another possible source of information is firm and industry
performance measures (e.g., earnings and sales figures), which would include past reported
amounts as well as future amounts forecasted by managers and analysts.
In addition, insiders face legal constraints and litigation concerns when trading or
entering trading plans while in possession of material non-public information. Under Rule 10b5,
promulgated by the U.S. Securities and Exchange Commission (SEC) under the Securities and
Exchange Act of 1934, insiders are prohibited from trading while in possession of material non-
public information. Enforcement of Rule 10b5 by the SEC has historically focused on insiders’
trades occurring prior to sharp price movements (Fried 1998), which might be expected around
goodwill impairments. During our sample period, insiders also can trade within Rule 10b5-1
trading plans (promulgated by the SEC in October 2000). Rule 10b5-1 trading plans allow
insiders to enter into preplanned trades while not in possession of material nonpublic
information. The intent of such plans is to provide insiders with a means to engage in
uninformed diversification trade. However, prior and concurrent research provides evidence
that, despite the assumption of uninformed diversification trade, insiders engage in strategic trade
within Rule 10b5-1 trading plans as well (Henderson, Jagolinzer, and Muller 2009; Jagolinzer
2009).
In summary, corporate insiders face incentives to sell their holdings prior to the public
release of goodwill impairments if such disclosures reveal managers’ private information. In
addition, if insider net selling activity reflects insiders’ private information regarding the
underlying impact of a future goodwill write-off, we expect firms whose insiders are net sellers
to experience relatively lower abnormal returns subsequent to insider trading. However,
11
managers will not trade prior to goodwill impairment announcements if market prices largely
reflect upcoming impairments due to investors having access to other sources of information
regarding goodwill impairments, or if managers face significant legal jeopardy. These arguments
lead to our two research hypotheses (stated in alternative form):
H1: Corporate insiders are net sellers of their company’s shares prior to goodwill
impairment announcements.
H2: For firms reporting goodwill impairments, firms where corporate insiders are net sellers
of their company’s shares prior to goodwill impairment announcements face lower abnormal returns relative to firms where corporate insiders are not net sellers of their company’s shares.
III. SAMPLE SELECTION AND DESCRIPTIVE INFORMATION
We first identify firms on Compustat Xpressfeed Quarterly reporting goodwill
impairments during the period 2002–2007. Our starting point is 2002 to coincide with the
effective year of SFAS 142 for most firms and to maintain consistency in the reporting standards
underlying goodwill impairment, since SFAS 142 introduced a number of significant changes for
goodwill valuation and reporting. Because some of our empirical tests focus on the time period
preceding the reported goodwill impairment, we retain only the first quarterly goodwill
impairment for each firm that reports multiple quarterly goodwill impairments, to avoid potential
confounding effects arising from overlapping pre-goodwill impairment periods. We further
restrict our sample to firms listed on the three major US stock exchanges (AMEX, NASDAQ and
NYSE) to ensure sufficient liquidity in the markets the insiders may be trading on. After
eliminating observations lacking necessary data for our analyses, we obtain a sample of 612
goodwill impairment firms (“GW Impair” sample). We collect financial information from
12
Compustat Xpressfeed Quarterly, stock returns from CRSP, and insider trading from Thompson
Financial.7
Panel A of Table 1 presents descriptive statistics for the GW Impair sample (N = 612) as
well as for all Compustat firm-quarter observations having positive goodwill values but no
reported goodwill impairments during our sample period (N = 49,574). The GW Impair
observations have higher representation in the fourth fiscal quarter. However, the final column
suggests the magnitude of goodwill impairments is not systematic across quarters, and ranges
from 1% to 23% of lagged market value of equity. Panel B presents the frequency of goodwill
impairments for our sample firms, with the majority of firms (63.6%) reporting only one
goodwill impairment over the six-year sample period. Table 2 provides the industry composition
of the GW Impair sample, which is generally similar to that of the Compustat population of firms
reporting goodwill, with the impairment sample reflecting a higher (lower) proportion of firms in
the electronic equipment (depository institutions) industry. The three industries with the largest
representation in our GW Impair sample are business services (13.4%), electronic equipment
(10.1%), and machinery and computers (6.2%).
Table 3 presents descriptive statistics comparing the GW Impair sample to the overall
sample of Compustat firms not reporting goodwill impairments. There is some evidence that
firms reporting goodwill impairments are smaller (SIZE) though slightly more followed by
analysts (FOLLOW). GW Impair firms have significantly lower profitability (ROA) and higher
book to market ratios (BTM), with the latter two variables inclusive of any reported goodwill
impairment. GW Impair firms also have relatively higher levels of goodwill in the quarter just
prior to the reported goodwill impairment (LAG_GW); however, this difference becomes
7 We are unable to collect information regarding the acquisitions that give rise either to the underlying goodwill or the reported goodwill impairment, as firms rarely provided sufficient disclosure to link the impairment to specific acquisitions.
13
insignificant after the impairment is reported (GW). The GW Impair firms also report more
frequent restructuring charges (RESTRUCT_D) and concurrent write-offs of non-goodwill assets
(OTH_IMPAIR_D). These latter differences reveal that goodwill impairments do not occur in
isolation, and suggest the importance of controlling for other concurrent charges in our
multivariate analyses.8 Finally, the mean (median) magnitude of the goodwill impairment for
our sample is 11.9% (2.3%) of the beginning market value of equity, indicating these
impairments are economically significant events.
To further assess the economic significance of goodwill impairment losses, we present in
Panel A of Figure 1 stock returns over the window (–30, 2) relative to day 0, the earnings
announcement date for the goodwill impairment quarter. Our choice of day –30 to begin the
trading window follows prior insider trading literature (Ke, Huddart, and Petroni 2003). In
addition, this window measurement is chosen to capture both the earnings announcement and
pre-announcement for the impairment quarter, since Bens, Heltzer, and Segal (2007) find that
most goodwill impairment announcements are included with either the quarterly earnings
announcement or the pre-announcement. We also decompose the sample of goodwill
impairment firms into two groups: those with low quality information environments, and those
with high quality information environments. This isolates a group of firms―those having low
quality information environments―in which early price discovery of goodwill impairments is
less likely to occur. We designate firms as having low (high) quality information environments
when analyst following is below (above) the median. This follows prior research, which
suggests that competition among analysts leads to more timely price discovery (e.g., Brennan,
8 The mean for restructuring charges scaled by lagged market value of equity for the sample of 612 goodwill impairment firms is 0.005, while that for other impairment charges similarly scaled is 0.010. Thus, while these charges occur frequently within the goodwill impairment sample, their relative economic magnitude appears considerably smaller than the goodwill impairment charges.
14
Jegadeesh and Swaminathan 1993; Brennan and Subramanyam 1995; Kimbrough 2007) and
lower information asymmetries in markets (e.g., Hong, Lim, and Stein 2000; Barth and Hutton
2004; and Ellul and Panayides 2008).
Panel A of Figure 1 presents descriptive information consistent with this notion. First,
there is a –5.53% (–2.96%) negative stock return surrounding the announcement of the goodwill
impairment for firms that operate in a low (high) quality information environment, as measured
over the trading window of (–30, 2). Untabulated analysis indicates that the magnitude of stock
returns over the window (–2, +2) relative to the earnings announcement for the goodwill
impairment quarter is on average –2.44% (–0.53%) for the low (high) quality information
environment firms. In addition, data presented below Panel A reveals that for low quality
information environment firms, a larger portion of the cumulative abnormal return occurs in
proximity to the announcement of the goodwill impairment. For example, 56.07% of the total
price response over the window (–375, 2) (which corresponds to approximately 18 calendar
months or 6 fiscal quarters) occurs in the window (–30, 2) for the low quality information
environment firms, relative to 11.84% for firms designated as having high quality information
environments. Taken together, these findings are consistent with goodwill impairments having
economic significance, especially for firms operating in a low quality information environment
where price discovery occurs closer to the accounting recognition of goodwill impairments.
To provide a more complete description of this price discovery process, Panel B of
Figure 1 reveals long-window stock returns before goodwill impairment announcements. Low
quality information environment firms exhibit relatively less negative run-downs in stock price
in the window (–375, –31), consistent with delayed price discovery of goodwill impairments for
this subset of firms. Data presented below this panel show that the low quality information
15
environment firms experience positive average earnings surprises for five out of the six quarters
preceding the goodwill impairment, while high quality information environment firms have
significantly negative earnings surprises. These differences highlight the importance of
controlling for unexpected earnings in our multivariate analysis. In addition, both panels suggest
that separate examination of these subsamples, with particular focus on firms having low quality
information environments, will better isolate a setting in which any identified price response
more likely reflects information relating to the goodwill impairment, versus other (non-goodwill
related) economic performance of the firm.
IV. RESEARCH DESIGN AND EMPIRICAL RESULTS
We now present the results of our empirical analyses examining insider trading behavior
preceding reported goodwill impairments. We first examine whether the incidence of net insider
selling is higher for firms reporting goodwill impairments relative to those not reporting
goodwill impairments. We then focus our analyses on those firms reporting goodwill
impairments, and examine whether the abnormal returns for firms with insiders that are net
sellers are more negative than for those firms with insiders that either do not trade or are net
buyers. Following our above discussion, we also separately conduct both analyses on two
subsamples of firms: those having low versus high quality information environments.
Insider Trading Before Goodwill Impairments
Based on the findings of prior research that the reporting of goodwill impairments lags
the economic impairment of goodwill (e.g., Hayn and Hughes 2006; Ramanna and Watts 2008),
we investigate whether informed investors, specifically firm insiders, take advantage of the
delays in the recognition of goodwill impairments and trade in anticipation of them. We
16
structure our multivariate analysis by focusing on the two years preceding the reported goodwill
impairment, divided into four non-overlapping six-month periods (see Figure 2 for a timeline).
We adopt this relatively long time frame based on the findings of prior research, which provides
evidence that firm insiders are able to disguise their informed trading by avoiding trades reported
too close to a relevant reporting event. For instance, Ke, Huddart, and Petroni (2003) document
that stock sales by insiders increase three to nine quarters prior to a break in a string of
consecutive quarterly earnings increases. In addition, a longer time frame appears particularly
relevant for our setting of goodwill impairments, as SFAS 142 requires firms to test their
goodwill for impairment annually (at the same time every year), suggesting insiders may obtain
private information about future impairments at least one year before the actual loss reporting. If
a firm narrowly avoids recording a goodwill write-off in the current period and the firm insiders
anticipate no substantial economic improvement over the following reporting period, they may
infer that a future goodwill write-off is unavoidable.9
To mitigate the concern that our empirical model captures trading activities that are
unrelated to the desire to profit from the anticipated goodwill impairment reporting, we estimate
our regressions using a comprehensive sample that includes both the goodwill impairment firms
and a group of control firms. The latter is comprised of all firms having available data for our
analyses, but not reporting goodwill impairments. Inclusion of these firms provides a benchmark
to control for typical insider trading activity coinciding with our sample period. Rather than
using all possible observations for each control firm, which would result in more than 120,000
control firm year observations, we sample the control firms at the same rate as the sample
firms—one observation per firm. Specifically, similar to the selection of comparison firms in
9 In addition, anecdotal evidence indicates that testing goodwill for impairment is a lengthy task than can last several months (PriceWaterhouseCoopers 2008).
17
related research (e.g., Lee, Mikkelson, and Partch 1992), we randomly select a single quarter
within our sample period to determine the reporting window for each control firm.
We first conduct univariate comparisons of the occurrence of net insider selling (denoted
as NETSELL_D) preceding the reported goodwill impairment. We measure NETSELL_D as an
indicator variable equal to one for firms that experience net insider selling (that is, total sales by
insiders that exceed total purchases), and zero otherwise. Three salient features of our net insider
selling variable warrant discussion.
First, we employ a dichotomous (versus continuous) measure as the functional form
relating net insider selling and our primary variable of interest (goodwill impairment) is unclear.
That is, a number of constraints (e.g., litigation concerns) likely inhibit substantial net insider
trading, suggesting net insider selling is not uniformly increasing in the existence of an
impairment.10
Second, we define insiders based on transactions of corporate officers.11 We include all
officers for two reasons. Senior officers (e.g., CEO, CFO) likely face greater scrutiny and thus
litigation risk, creating tension on whether observed trading will occur only for this subset of
officers. In addition, it is difficult to isolate what level of officer possesses private information
regarding the value―including potential impairment―of goodwill. However, inclusion of
officers lacking this information will lead to noise in the construction of NETSELL_D. Finally,
10 Nonetheless, we also compute a continuous measure of insider trading defined as the number of shares sold divided by the number of shares sold plus the number of shares purchased by firm insiders. Similar measures have been used in prior insider trading studies (e.g., Lakonishok and Lee 2001; Piotroski and Roulstone 2005). In our setting, about 60% of the goodwill impairment firms have insider sale ratios equal to zero (i.e., no selling occurs), 35% equal to one (i.e., all trades are sales), and only about 5% of the observations fall between zero and one. The large number of observations equal to one is expected, which is attributable to very few insider purchases for the impairment firms. Not surprisingly, our univariate results are unchanged employing this alternative measure. Results are also similar using an insider sale ratio based on trade values (as opposed to number of shares). 11 The SEC defines the term “officer” to include: a company’s president; principal financial officer; principal accounting officer; vice president in charge of a principal business unit, division, or function (such as sales, administration, or finance); and any other person, who performs a policy-making function for the company. Thus, a vice president, who is not in charge of a principal unit and is not a policy-maker, is not considered an insider or Section 16 officer.
18
we exclude transactions of company directors, since prior research (e.g., Piotroski and Roulstone
2005) hypothesizes and finds that directors only meet periodically and are less likely than firm
officers to have timely access to performance-related information.
Third, following prior literature (Piotroski and Roulstone 2005; McVay et al. 2006), we
include in the calculation of our insider trading measure only open-market transactions to
eliminate the impact of insider trades related to option and stock grants. We assess NETSELL_D
over the following four non-overlapping windows, all defined with respect to month 0, the last
month of the quarter in which the goodwill impairment is reported: (–23,–18), (–17,–12), (–11,–
6), and (–5, 0).12
To examine the cross-sectional determinants of insider trading, we then adopt the
following logistic regression to investigate whether insiders strategically sell ahead of anticipated
goodwill impairments:13
( ) 0 1 2 3 4
5 6 7,
0
_ _
_ _ _ _
ϕ ϕ ϕ ϕ ϕ
ϕ ϕ ϕ ε− −
=
= + + + +
+ + + +∑
W indow
K
t k t k
k
NET SELL D GW I D SIZE BT M RET BH
REST RUCT D OT H IMPA IR D UE Q (1)
where: NETSELL_D(Window) = an indicator variable equal to one for firms that experience net insider
selling (i.e., total sales greater than total purchases) over the indicated six-month trading window, and zero otherwise;
GWI_D = an indicator variable equal to one for firms reporting a goodwill
impairment in quarter t, and zero otherwise; SIZE = log of total assets measured at the end of the indicated six-month
window; BTM = the book-to-market ratio measured at the end of the indicated six-month
12 Following prior literature (Lakonishok and Lee 2001), we aggregate insider trades over six-month windows. Calculating insider trading measures based on shorter periods (such as a month or a quarter) results in many companies having no trades. Lakonishok and Lee (2001) argue that insider trading measures computed over six-month windows have greater predictive power. 13 We also estimate equation (1) including time fixed effects to control for potential effects specific to different quarters within our sample period. The results are very similar to those tabulated.
19
window; RETBH = the buy-and-hold return measured over the six-month window prior to
the indicated insider trading measurement window; RESTRUCT_D = an indicator variable equal to one for firms reporting restructuring
charges in quarter t (the goodwill impairment quarter), and zero otherwise;
OTH_IMPAIR_D = an indicator variable equal to one for firms reporting any non-goodwill
impairments in quarter t (the goodwill impairment quarter), and zero otherwise; and
UE_Qt-k = the unexpected earnings for quarter t-k, where the unexpected earnings
are measured as earnings at the end of quarter t-k minus earnings at the end of quarter t-k-4, scaled by market value of equity at the end of quarter t-k-1.
Paralleling the univariate analysis, equation (1) is estimated separately for four dependent
variables, NETSELL_D(Window), measured over the following monthly intervals preceding the
reported impairment: (–23,–18), (–17,–12), (–11,–6), and (–5,0). Our primary variable of
interest is GWI_D. Similar to NETSELL_D, we measure goodwill impairments as a dichotomous
variable. We do not use a continuous variable as it would require the market expectation of
goodwill impairment (which is not separately observable), and as insider trading may not
increase monotonically with the size of the unexpected goodwill impairment (e.g., due to
possible litigation risks). If firm insiders are able to predict forthcoming goodwill impairments,
and trade in anticipation of these reported amounts, we expect the probability that a firm
experiences net informed selling to be higher for the goodwill impairment firms relative to the
non-impairment firms, leading to a positive predicted coefficient for GWI_D and our primary test
of H1.
We also include variables to control for other determinants of insider trading. First, we
include firm size (SIZE) to proxy for portfolio rebalancing that is unrelated to private information
20
trading. For larger firms, insiders are likely to hold larger stock and stock option positions.
Therefore, we expect a positive association between the probability of selling stock and SIZE, if
insiders with large stock holdings sell in an attempt to reduce their exposure to firm idiosyncratic
risk. We also include the book-to-market ratio (BTM) to proxy for growth, with both
components measured at the end of the indicated six-month period. Prior literature (Rozeff and
Zaman 1988; Piotroski and Roulstone 2005) shows that insiders are more likely to hold larger
stock positions as stocks change from growth to value categories; thus, we predict a negative
coefficient for BTM. In addition, we include prior period returns (RETBH) because insiders tend
to be contrarian (Lakonishok and Lee 2001), leading to a predicted positive coefficient.
We also include control variables, RESTRUCT_D and OTH_IMPAIR_D, for other types
of restructuring and impairment charges to address the concern that we attribute any potential
informed trading activity to goodwill impairments when, in fact, such trading may reflect
information contained in other types of charges. Because the expected direction of trading in the
context of these other charges is unclear, we do not predict the sign for these coefficients.
Finally, to control for informed trading that could occur on a broader set of financial data,
we also include in equation (1) the unexpected earnings for each of the quarters occurring from
the start of the trading window up to the quarter of the reported impairment. Specifically,
following Ke, Huddart, and Petroni (2003), we define UE_Qt-k as the unexpected earnings for
quarter t-k, measured as the seasonally-adjusted change in quarterly earnings scaled by the
lagged market value of equity.14 We include K quarters of unexpected earnings for each window
to capture the full series of reporting quarters overlapping with the trading windows, where K is
14 In sensitivity analyses, we investigate whether our findings are robust to breaks in earnings strings. Following the findings of Ke, Huddart, and Petroni (2003), we include an indicator variable equal to one for a break in the string of earnings increases for the impairment quarter for strings of 3 quarters (and alternatively, a string of 5 quarters), and zero otherwise. Our results remain unchanged.
21
the number of quarters between the insider trading window and the impairment reporting quarter.
For example, when the insider trading variable is measured over the window months (–5, 0), we
include in equation (1) unexpected earnings for two quarters—the goodwill impairment quarter
and the quarter prior to the goodwill impairment. When the insider trading variable is measured
over the window months (–11, –6), we include in equation (1) unexpected earnings for four
quarters—the goodwill impairment quarter and the three quarters prior to the goodwill
impairment. Finally, note that the unexpected earnings for the quarter with the reported goodwill
impairment excludes the goodwill and other asset impairments and restructuring charges from its
measurement, as these are all explicit variables within the regression.
Table 4 presents the univariate analysis of insiders’ net selling behavior prior to the
impairment announcement. We compare the mean and median values of NETSELL_D for the
goodwill impairment firms versus the non-impairment (control) firms for each of the four
windows. For three out of the four six-month trading windows, firms reporting a goodwill write-
off are significantly more likely to experience net insider selling relative to the non-impairment
firms. Specifically, we find that 43.5%, 46.8% and 42.9% of the goodwill impairment firms
experience net insider selling over the corresponding windows of (–23,–18), (–17,–12) and (–
11,–6) prior to the reported goodwill impairment. Over the same measurement windows, only
36.4%, 36.4% and 37.1% of non-impairment firms experience net insider selling. The
differences between these two firm types are significant at a p-value less than 1% both for the
means and the medians. However, we fail to find evidence of differences in net insider selling
for the window (–5, 0).
Table 5 presents multivariate results from estimating equation (1) for each of the four
windows. As insiders’ selling activities may be correlated in time, we report standard errors
22
clustered by time (Petersen 2009). We do not cluster the standard errors by firm, since each firm
in our analysis enters the sample just once. For the windows (–17,–12) and (–11,–6), consistent
with our hypothesis and with the univariate analysis, we find a significantly positive coefficient
for GWI_D (z-statistics = 3.115 and 2.177, respectively). This finding indicates that, controlling
for other determinants of net insider selling activity, firms reporting goodwill impairments
experience significantly higher insider selling over the windows (–17,–12) and (–11,–6)
compared with firms reporting no such impairments. We also find some evidence (although
marginally significant) of higher insider selling for goodwill impairment firms over the window
(–23, –18). However, when insider trading activities are measured over the window (–5, 0), we
fail to find evidence of significantly higher insider selling activity as the coefficient on GWI_D is
insignificant (z-statistics = 0.117). Among the control variables, we find that the coefficients for
SIZE, BTM, and RETBH are significantly positive, negative, and positive (respectively) in the
predicted direction across all trading windows, as predicted. We find limited significance across
the windows for the remaining control variables.
It is interesting to note that the timing of abnormally high insider selling volume centers
near the (likely) annual impairment testing for the year preceding the actual goodwill impairment
reporting (i.e., the 18 to 6 months prior to the reported goodwill impairment). While not
definitive, this evidence is suggestive that during the current year’s impairment testing insiders
are able to generate private information about the probability of having to record goodwill
impairments for the ensuing year, and trade in accordance with this inferred probability.15
We then examine insider trading on the subsamples of firms having low versus high
quality information environments. As discussed previously, we designate firms as having low
15 For example, a narrow avoidance of a goodwill impairment loss for the current year may allow insiders to infer that absent a substantial improvement in the firm’s economic performance, a write-off is unavoidable for the next year.
23
(high) quality information environments if they have below (above) median analyst following.
For this analysis, our primary interest resides in the low quality information environment, as this
will provide evidence that our findings are robust when news regarding goodwill impairment
charges occurs closer to its formal release (i.e., to mitigate the likelihood that trading on non-
goodwill information explains our findings).
For this insider trading analysis, analyst following is determined at the end of each of the
four trading windows.16 Table 6 presents the results of this subsample analysis. Panel A, which
focuses on firms with low quality information environments, reveals that insiders exhibit
abnormal selling across all four trading windows, reflected in significantly positive coefficients
for GWI_D for the trading window –23 to –18 (z-statistic = 3.496), the trading window –17 to –
12 (z-statistic = 6.962), the trading window –11 to –6 (z-statistic = 3.040), and the trading
window –5 of 0 (z-statistic = 2.948). In contrast, results in Panel B, which focus on firms with
high quality information environments, provide weaker evidence of abnormal selling, reflected
in an insignificant coefficient for GWI_D for the trading window –23 to –18 (z-statistic = 1.322),
a significant coefficient for the trading window –17 to –12 (z-statistic = 2.288), and insignificant
coefficients for the trading windows –11 to –6 (z-statistic = 1.498) and –5 of 0 (z-statistic = –
0.770). Results for the control variables across both panels are unchanged from the primary
analysis. Overall, the results are consistent with insiders exhibiting abnormal selling preceding
the announcement of goodwill impairments, and that―at least for firms designated as having
low quality information environments―this is unlikely driven by economic performance
16 Because the distribution of analyst following contains only integers, a relatively large number of observations have analyst following values equal to the median value. In our main analysis we allocate these observations to the low quality information environment. We also conduct an additional sensitivity test to provide an equal number of observations for each subsample. Specifically, for firms with equivalent analyst following values at the median, we designate those with smaller (larger) market capitalization as having lower quality (higher quality) information environments, under the assumption that market capitalization is positively correlated with overall quality of the information environment. Our inferences remain unchanged.
24
unrelated to goodwill impairments.
Abnormal Returns to Insider Trades Preceding Goodwill Impairments
The evidence presented in Tables 4 through 6 indicates that insider selling for firms
reporting goodwill impairments exceeds that for non-impairment firms. If this abnormally high
insider selling activity is based on insiders’ private information about the negative effect of a
future goodwill write-off, we expect firms where insiders are net sellers to experience lower
abnormal returns subsequent to insider trading compared to firms where insiders are net buyers
or do not trade. To investigate this possibility, we first examine univariate differences in
abnormal returns across these two groups of firms. We compute the abnormal buy-and-hold
returns over windows that start directly after the end of six insider trading windows, and end
with the last month of the goodwill impairment quarter (i.e.., month 0). The six windows are as
follows:
Insider Trading Window Months Abnormal Buy-and-Hold Return Months
( –8, –3) ( –2, 0) (–11, –6) ( –5, 0) (–14, –9) ( –8, 0) (–17, –12) (–11, 0) (–20, –15) (–14, 0) (–23, –18) (–17, 0)
To examine the cross-sectional determinants of abnormal buy-and-hold returns associated
with insider trades for firms reporting goodwill impairments, we adopt the following equation:17
( ) 0 1 2 3 4
5 6 7,
0
_ _
_ _ _ _
φ φ φ φ φ
φ φ φ ς− −
=
= + + + +
+ + + +∑
W indow
K
t k t k
k
A BNOR MA L RET NET SELL D SIZE BT M RET BH
REST R UCT D OT H IMPA IR D UE Q(2)
where:
17 We also estimate equation (2) for all of our six windows of interest including time fixed effects to control for potential effects specific to different quarters within our sample period, obtaining similar results.
25
ABNORMAL_RET = the buy-and-hold abnormal returns to insider trades, computed over
windows starting directly after the end of the insider trading window and ending with the last month of the goodwill impairment quarter (month 0).
The dependent variable, ABNORMAL_RET, is measured as the difference between the raw return
for the firm and the corresponding return of the value-weighted market index. All other variables
are as previously defined. We measure insider trades over six-month rolling windows starting
two years prior to the impairment event, and identify goodwill impairment firms where insiders
are net sellers (NETSELL_D = 1) versus goodwill impairment firms where insiders are either net
buyers or do not trade (NETSELL_D = 0).18 We then compute abnormal buy-and-hold returns
over the period beginning directly after the insider trading window and ending with the goodwill
impairment quarter. Thus, paralleling the univariate analyses, equation (2) is estimated
separately for six dependent variables, ABNORMAL_RET(Window) measured over the following
monthly intervals preceding the reported impairment: (–17,0), (–14,0), (–11,0), (–8,0), (–5,0),
and (–2,0).
Our primary variable of interest is NETSELL_D. The variable captures the market-
adjusted return avoided by insiders, who sell stock during the six-month insider trading window
rather than waiting to execute the sale until the goodwill impairment is publicly announced. The
variable is defined similarly for firms reporting goodwill impairments but not having insiders
with net selling activity. If firm insiders selling their company’s shares prior to goodwill
impairment announcements have greater private information regarding negative circumstances
leading to an economic impairment of goodwill than those not selling their company’s shares
prior to the announcement, then we expect firms with insiders selling their shares to experience
18 In the abnormal return to insider trading analysis, we compare goodwill impairment firms where insiders are net sellers versus all the other goodwill impairment firms where insiders are either net buyers or do not trade. We combine net buyers with non-traders because in our sample the number of net buyers is very small.
26
lower abnormal returns. This is our primary test of H2.
To address the concern that the differences in buy-and-hold returns between firms with
and without net insider selling are, in fact, driven by other firm characteristics that are correlated
with insider trading, we include in equation (2) control variables for firm size (SIZE), book to
market ratio (BTM) and past returns (RETBH). As explained previously in the variable
description for equation (1), these firm characteristics should be correlated with insider trading.
In addition, prior literature identifies these characteristics as stock return determinants. We also
include control variables for other types of impairments and restructuring charges
(RESTRUCT_D and OTH_IMPAIR_D) and for the unexpected earnings for each of the quarters
occurring from the start to the end of the return window (UE_Qt-k). These variables address the
potential concern that we attribute the abnormally low returns for net insider selling impairment
firms to insiders’ private information about the loss in value of goodwill when, in fact, the lower
returns are driven by a higher frequency of restructuring charges, other types of impairments,
and/or negative earnings surprises for the net insider selling firms relative to no trading or net
buying firms.
Table 7 presents univariate mean and median buy-and-hold abnormal return comparisons
for goodwill impairment firms having net insider selling versus all other goodwill impairment
firms. Figure 3 plots the mean buy-and-hold abnormal returns for each window. All goodwill
impairment firms experience negative average abnormal returns in the quarters leading to the
impairment, reflected in the negative mean and median returns for both net insider selling
impairment firms as well as all other impairment firms across all return windows. However, the
firms where insiders are net sellers have significantly more negative abnormal stock returns
compared to the firms where insiders are not net sellers. Specifically, significant negative
27
differences occur for the return windows of –11 to 0 (where the incremental negative return is –
0.070, t-statistic = 2.020), –14 to 0 (–0.108, t-statistic = 2.740), and –17 to 0 (–0.129, t-statistic =
–3.010). The difference between net insider selling and the other impairment firms are not only
statistically significant, but also economically significant. For example, using the insider trading
window of (–23,–18), firms where insiders are net sellers experience an average abnormal return
of –23.4% over the eighteen months leading to the reported goodwill impairment, compared to
only –10.5% for the impairment firms where insiders are not net sellers.
Table 8 presents multivariate results from estimating equation (2) for each of the six
windows. Again, as insiders’ selling activities may be correlated in time, we report standard
errors clustered by time (Petersen 2009). For the return windows (–17, 0), (–14, 0), and (–11, 0),
consistent with our hypothesis and with the univariate analysis, we find a significantly negative
coefficient for NETSELL_D (t-statistics = –2.270, –2.380, –2.290, respectively). This finding
indicates that, controlling for other determinants of abnormal returns, insiders selling their
company’s shares prior to goodwill impairment announcements have more severe negative
private information than those not selling their company’s shares prior to the announcement.
However, when returns to insider trading activities are measured over the windows (–8, 0), (–5,
0), and (–2, 0), we fail to find evidence of relatively lower abnormal returns as the coefficient on
NETSELL_D is insignificant (t-statistics = –0.780, –0.780, and –0.870, respectively). Among the
control variables, we find that SIZE, RETBH, and OTHER_IMPAIR_D are significant across
most trading windows, as predicted. We find limited significance across the windows for the
other control variables.
Paralleling our previous results, we examine abnormal returns for the two subsamples of
firms having low quality versus high quality information environments. As shown in Panel A of
28
Table 9, our results for the variable NETSELL_D are robust, albeit weaker, for the subsample of
firms having low quality information environments. Again, this setting more likely reflects
insider trading relating to the goodwill impairment, versus other (non-goodwill related)
economic performance of the firm. Panel B reveals generally insignificant coefficients for
NETSELL_D for the subsample of firms having high quality information environments. Taken
together, these findings suggest that firm insiders are able to profit from their private information
regarding goodwill impairments by timing their trades in advance of its incorporation by the
equity market or recognition by the firm’s accounting system.
V. CONCLUSION
This study investigates whether insiders strategically sell their shares prior to goodwill
impairments. We examine insider trading in this setting to provide evidence of a specific
component of earnings that insiders are able to trade ahead of—i.e., to provide evidence of a
specific source of insiders’ private information. We select the setting of goodwill impairments
for three reasons. Goodwill impairments tend to be economically large, averaging 11.9% of the
market value of equity during our sample period of 2002-2007. In addition, managers likely
have material private information regarding future cash flow estimates (which relate to the value
of recorded goodwill) through their internal budgeting processes. Finally, managers’ private
information advantage may be relatively long-lived due to goodwill impairment testing rules that
may delay the accounting recognition of economic goodwill impairments. Whether insiders
trade in anticipation of reported goodwill impairments is unclear, as equity markets may
anticipate economic impairments of goodwill prior to the reported impairment, and governance
and litigation constraints may inhibit such strategic trading behavior.
Our empirical analyses reveal that insiders of firms reporting goodwill impairments
29
exhibit higher net selling behavior relative to firms not reporting goodwill impairments up to
twenty-four months prior to the goodwill impairment announcement. Further analysis reveals
that among firms reporting goodwill impairments, those with net selling among insiders exhibit
significantly more negative abnormal stock returns relative to those not reporting net selling
among insiders. Results across both sets of analyses are robust to inclusion of controls for other
determinants of insider trading activity and abnormal stock returns. Further, these results are
robust to examining a subsample of firms having a relatively low quality information
environment—where price discovery regarding goodwill impairments is significantly delayed.
This setting provides evidence in an environment where strategic insider trading behavior can be
more directly linked to information regarding goodwill impairments, as the price discovery
regarding the goodwill impairment occurs closer to its formal release.
Overall, these results build upon prior insider trading research by providing evidence of a
specific reporting item about which insiders are able to strategically trade prior to its full
discovery by the equity market and its recognition within the financial statements. In addition,
these results build upon prior research investigating managers’ discretion over goodwill
impairments by providing evidence of a managerial incentive to delay the accounting recognition
of goodwill impairments. Finally, these results build on prior research investigating the market
reaction to the accounting recognition of goodwill impairments by providing evidence that the
market reaction is substantially greater in low quality information environments.
30
REFERENCES
ACCOUNTING PRINCIPLES BOARD. Opinion No. 17: Intangible Assets (1970). New York NY.
AHMED, A.; L. GULER. “Investors' Valuation of Goodwill and Goodwill Write-offs: Pre-
SFAS 142 versus Post-SFAS 142.” Working paper, Texas A&M University, 2007. ANANTHARAMAN, D. “Has SFAS 142 Diminished the Reliability of Goodwill Writeoffs?”
Working paper, Columbia Business School, 2007. BARTH, M.; AND A. HUTTON. “Analyst Earnings Forecast Revisions and the Pricing of
Accruals.” Review of Accounting Studies 9 (2004): 59-96. BEATTY, A.; AND J. WEBER. “Accounting Discretion in Fair Value Estimates: An
Examination of SFAS 142 Goodwill Impairments.” Journal of Accounting Research 44 (2006): 257–288.
BENEISH, D; AND M. VARGUS. “Insider Trading, Earnings Quality, and Accrual
Mispricing.” The Accounting Review 77 (2002): 755-791. BENS, D.; W. HELTZER; AND B. SEGAL. “The Information Content of Goodwill
Impairments and the Adoption of SFAS 142.” Working paper, University of Arizona, 2007. BRENNAN, M.; N. JEGADEESH; AND B. SWAMINATHAN. “Investment Analysis and the
Adjustment of Stock Prices to Common Information.” The Review of Financial Studies 6 (1993): 799-824.
BRENNAN, M; AND A. SUBRAHMANYAM. “Investment Analysis and Price Formation in
Stock Markets.” Journal of Financial Economics 38 (1995): 361-381. ELLIOT, J.; D. MORSE; AND G. RICHARDSON. “The Association between Insider Trading
and Information Announcements.” Rand Journal of Economics 15 (1984): 521-536. ELLUL, A; AND M. PANAYIDES. “Do Financial Analysts Restrain Insiders’ Informational
Advantage?” Working paper, Indiana University and University of Utah (2008). FINANCIAL ACCOUNTING STANDARDS BOARD. Statement of Financial Accounting
Standards No. 121, Accounting for the Impairment of Long-Lived Assets and for Long-Lived
Asset to Be Disposed of (1995). FASB, Norwalk, CT. FINANCIAL ACCOUNTING STANDARDS BOARD. Statement of Financial Accounting
Standards No. 131, Disclosures about Segments of an Enterprise and Related Information
(1997). FASB, Norwalk, CT.
31
FINANCIAL ACCOUNTING STANDARDS BOARD. Statement of Financial Accounting
Standards No. 142, Goodwill and Other Intangible Assets (2001b). FASB, Norwalk, CT. FRANCIS, J.; D. HANNA; AND L. VINCENT. “Causes and Effects of Discretionary Asset
Write-Offs.” Journal of Accounting Research 34 (1996): 117–134. FRIED, J. “Reducing the Profitability of Corporate Insider Trading Through Pretrading
Disclosure.” Southern Californian Law Review 71 (1998): 303–392. GIVOLY, D.; AND D. PALMON. “Insider Trading and the Exploitation of Inside Information:
Some Empirical Evidence.” Journal of Business 58 (1985): 69–87. GU, F.; AND B. LEV. “Overpriced Shares, Ill-Advised Acquisitions, and Goodwill
Impairment.” Working paper, New York University, 2008. HAYN, C.; AND P. J. HUGHES. “Leading Indicators of Goodwill Impairment,” Journal of
Accounting, Auditing and Finance 21 (2006): 223–265. HENDERSON, T.; A. JAGOLINZER; AND K. MULLER. “Strategic Disclosure of 10b5-1
Trading Plans.” Working paper, University of Chicago, 2009. HLOUSEK, R. “Asset Impairment: Transitioning under SFAS 142.” Houlihan Valuation
Advisors’ Newsletter, Spring 2002. Available at: www.cogentvaluation.com HONG, H.; T. LIM; AND J. STEIN. “Bad News Travels Slowly: Size, Analyst Coverage, and
the Profitability of Momentum Strategies.” Journal of Finance 55 (2000): 265-295. HUDDART, S.; B. KE; AND C. SHI. “Jeopardy, Non-Public Information, and Insider Trading
Around SEC 10-K and 10-Q Filings.” Journal of Accounting & Economics 43 (2007): 3-36. JAGOLINZER, A. D. “SEC Rule 10b5-1 and Insiders’ Strategic Trade.” Management Science
55 (2009): 224-239. JAGOLINZER, A.; AND D. ROULSTONE. “The Effects of Insider Trading Regulation on
Trade Timing, Litigation Risk, and Profitability.” Working paper, Stanford University and Ohio State University (2008).
JAYARAMAN, S. “Earnings Volatility, Cash Flow Volatility, and Informed Trading.” Journal
of Accounting Research 46 (2008): 809-851. JOHN, K.; AND L. LANG. “Strategic Insider Trading Around Dividend Announcements:
Theory and Evidence.” Journal of Finance 46 (1991): 1361–1398. KARPOFF, J.M.; AND D. LEE. “Insider Trading Before New Issue Announcements.”
Financial Management 20 (1991): 18–26.
32
KE, B.; S. HUDDART; AND K. PETRONI. “What Insiders Know About Future Earnings and How They Use It: Evidence From Insider Trades.” Journal of Accounting and Economics 35 (2003): 315–346.
KIMBROUGH, M. “The Influences of Financial Statement Recognition and Analyst Coverage
on the Market’s Valuation of R&D Capital.” The Accounting Review 82 (2007): 1195-1225. LAKONISHOK, J.; AND I. LEE. “Are Insider Trades Informative?” Review of Financial
Studies 14 (2001): 79–111. LEE, D.S.; W.H. MIKKELSON; AND M.M. PARTCH. “Managers’ Trading Around Stock
Repurchases.” Journal of Finance 47(1992): 1947–1961. LI, Z.; P.K. SHROFF; AND R. VENKATARAMAN. “Goodwill Impairment Loss: Causes and
Consequences.” Working paper, University of Minnesota, 2004. MCVAY, S; V NAGAR; AND VW TANG. “Trading Incentives to Meet the Analyst Forecast.”
Review of Accounting Studies 11 (2006): 575–598. PETERSEN, M. A. “Estimating Standard Errors in Finance Panel Data Sets: Comparing
Approaches.” Forthcoming, Review of Financial Studies. PIOTROSKI, J. D.; AND D. T. ROULSTONE. “Do Insider Trades Reflect both Contrarian
Beliefs and Superior Knowledge about Future Cash Flow Realizations.” Journal of
Accounting and Economics 39 (2005): 55-81. PRICEWATERHOUSECOOPERS. “Mergers and Acquisitions: A Snapshot.” Accessed from
www.cfodirect.com. December, 2008. RAMANNA, K.; AND R. WATTS. “Evidence from Goodwill Non-Impairments on the Effects
of Unverifiable Fair-Value Accounting.” Working Paper, Harvard University, 2008. RIEDL, E. “An Examination of Long-Lived Asset Impairments.” The Accounting Review
(2004): 823-852. ROGERS, J. “Disclosure Quality and Management Trading Incentives.” Journal of Accounting
Research 46 (2008): 1265-1296. ROZEFF, M. S.; AND M. A. ZAMAN. “Overreaction and Insider Trading: Evidence from
Growth and Value Portfolios.” Journal of Finance 53 (1998): 701–716. ROULSTONE, D. “Insider Trading and the Information Content of Earnings Announcements.”
Working paper, University of Chicago, 2006. SEYHUN, N. “Do Bidder Managers Knowingly Pay Too Much For Target Firms?” Journal of
Business 63 (1990): 439–464.
33
SEYHUN, N.; AND M. BRADLEY. “Corporate Bankruptcy and Insider Trading.” Journal of
Business 70 (1997): 189–216. SIVAKUMAR, K.; AND G. WAYMIRE. “Insider Trading Following Material News Events:
Evidence from Earnings.” Financial Management 23 (1994): 23–32.
FIGURE 1
Abnormal Returns for Firms Reporting Goodwill Impairments: Low-Quality versus High Quality Subsamples
Ratio of: (CAR over window –30, 2) / (CAR over indicated window) Unexpected
Earnings High quality
information environment Low quality
information environment
Total CAR window
High quality information environment
Low quality information environment
Mean t-statistic Mean t-statistic
Mean Mean UE_Q0 4.78% 4.45 *** 2.93% 1.38 (–375, 2) 11.84% 56.07% UE_Q-1 –0.64% –1.91 * 0.70% 0.70
UE_Q-2 –0.37% –1.95 * –0.32% –0.18 (–251, 2) 12.78% 36.20% UE_Q-3 –0.28% –1.14 0.35% 0.49
UE_Q-4 –0.41% –1.80 * 1.11% 0.96 (–126, 2) 23.42% 41.62% UE_Q-5 –0.24% –1.63 * 0. 31% 0.59
This figure presents the cumulative abnormal returns (CAR) for firms reporting goodwill impairments. Amounts for two sub-samples are shown: those having high quality information environments (defined as above median analyst following, represented with the dotted line), and those having low quality
-0.2
-0.13
-0.06
0.01
0.08
-37
5
-36
0
-34
5
-33
0
-31
5
-30
0
-28
5
-27
0
-25
5
-24
0
-22
5
-21
0
-19
5
-18
0
-16
5
-15
0
-13
5
-12
0
-10
5
-90
-75
-60
-45
Panel B: CAR over the window (-375,-31) for
goodwill impairment firms with high versus low
quality information environment
High Quality Info Environment Low Quality Info Environment
-0.056
-0.046
-0.036
-0.026
-0.016
-0.006
0.004
-30 -28 -26 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2
Panel A: CAR over the window (-30,2) for goodwill
impairment firms with high versus low quality
information environment
High Quality Info Environment Low Quality Info Environment
35
information environments (defined as below median analyst following, represented with the solid line). Panel A presents CAR for the window (–30, 2); Panel B presents CAR for the window (–375, –31); for both panels, day 0 is the announcement of the goodwill impairment. The trading windows of (–375, 2), (–251, 2) and (–126, 2) correspond to approximately 18, 12 and 6 calendar months, or 6, 4, and 2 fiscal quarters, respectively.
36
FIGURE 2
Timeline for the measurement of informed trading over the pre-goodwill impairment disclosure period
End of the GW Impairment
Fiscal Quarter
Window
(-11,-6)
Window
(-5,0)
Month 0
Window
(-17,-12)
Window
(-23,-18)
This figure presents a timeline for the measurement of the variables used in the informed trading analysis. The informed trading and the time-variant control variables are measured over four non-overlapping windows covering the two-year period prior to the reported goodwill impairment. The first window is measured over months –23 to –18 (inclusive); the second is measured over months –17 to –12 (inclusive); the third is measured over months –11 to –6 (inclusive); and the last window over months –5 to 0 (inclusive), where month 0 is the last month of the goodwill impairment fiscal quarter. For example, for a company that reports a goodwill impairment for the quarter ending December 31, 2006, the first window is measured from January 1, 2005 to June 30, 2005; the second window from July 1, 2005 to December 31, 2005; and so on.
Abnormal Buy-and-Hold Returns for Firms Reporting SFAS 142 Goodwill Impairments
This figure presents evidence that firms reporting goodwill impairments and with net insider selling transactions face relatiabnormal buy-and-hold returns as compared to firms reporting goodwill impairments and not exhibiting net insiderabnormal buy-and-hold returns to insider trades are computed over windows starting directly after the end of the insider trading window and
-25.0%
-20.0%
-15.0%
-10.0%
-5.0%
0.0%
(-17,0)
Firms with net insider selling -23.4%
All other -10.5%
Ab
no
rma
l b
uy
-an
d-h
old
re
turn
FIGURE 3
Hold Returns for Firms Reporting SFAS 142 Goodwill Impairments
This figure presents evidence that firms reporting goodwill impairments and with net insider selling transactions face relatihold returns as compared to firms reporting goodwill impairments and not exhibiting net insiderhold returns to insider trades are computed over windows starting directly after the end of the insider trading window and
(-14,0) (-11,0) (-8,0) (-5,0) (-2,0)
-22.3% -20.4% -16.8% -13.8% -8.4%
-11.5% -13.4% -15.8% -12.0% -8.0%
Months relative to goodwill impairment disclosure
This figure presents evidence that firms reporting goodwill impairments and with net insider selling transactions face relatively more negative hold returns as compared to firms reporting goodwill impairments and not exhibiting net insider selling transactions. The hold returns to insider trades are computed over windows starting directly after the end of the insider trading window and
38
ending with the last month of the goodwill impairment quarter (i.e., month 0). Thus, the first column presents results for insider trading calculated over months (–23, –18), and abnormal returns calculated over months (–17, 0).
TABLE 1
Descriptive Statistics of SFAS 142 Goodwill Impairments
Panel A: Goodwill impairments by reporting period, 2002–2007
Firms Reporting Goodwill
Impairments (a)
% of Total
Impairment Sample
(b)
Compustat Firms
Reporting Goodwill
(c)
% of Total Compustat Firms with Goodwill
(d)
Impairment Firms /
Compustat Firms
(a)/(c)=(e)
Mean of Reported
Impairment / Market Valuet-1
(f)
2002 Q1 3 0.5% 1,135 2.3% 0.26% 3% Q2 19 3.1% 1,507 3.0% 1.26% 17% Q3 28 4.6% 1,646 3.3% 1.70% 23% Q4 103 16.8% 1,762 3.6% 5.85% 19%
2003 Q1 7 1.1% 1,945 3.9% 0.36% 3% Q2 24 3.9% 1,814 3.7% 1.32% 13% Q3 21 3.4% 1,835 3.7% 1.14% 15% Q4 52 8.5% 1,924 3.9% 2.70% 14%
2004 Q1 9 1.5% 2,211 4.5% 0.41% 3% Q2 14 2.3% 2,087 4.2% 0.67% 7% Q3 25 4.1% 2,167 4.4% 1.15% 6% Q4 52 8.5% 2,243 4.5% 2.32% 7%
2005 Q1 8 1.3% 2,352 4.7% 0.34% 3% Q2 16 2.6% 2,227 4.5% 0.72% 9% Q3 13 2.1% 2,280 4.6% 0.57% 14% Q4 57 9.3% 2,358 4.8% 2.42% 10%
2006 Q1 12 2.0% 2,419 4.9% 0.50% 2% Q2 18 2.9% 2,252 4.5% 0.80% 8% Q3 16 2.6% 2,290 4.6% 0.70% 10% Q4 63 10.3% 2,380 4.8% 2.65% 13%
2007 Q1 8 1.3% 2,411 4.9% 0.33% 1% Q2 21 3.4% 2,247 4.5% 0.93% 8% Q3 15 2.5% 2,243 4.5% 0.67% 7% Q4 8 1.3% 1,839 3.7% 0.44% 4%
612 100.0% 49,574 100.0% 1.26%
Panel B: Goodwill impairments per firm, 2002–2007
Impairments Reported
During Sample Period
1
2
3
4
5
6
7 +
Total
Obs
Number of firms 389 139 47 17 8 5 7 612
Percent of sample 63.6% 22.7% 7.7% 2.8% 1.3% 0.8% 1.1% 100.0%
This table presents descriptive statistics for the sample of firms reporting goodwill impairments during the period 2002–2007, as well as comparative amounts for all other Compustat firms reporting positive goodwill but no goodwill impairment. Panel A presents the distribution by fiscal quarter. Panel B presents the distribution of sample firms by the number of reported goodwill impairments within the sample period. Only the first reported impairment for the 2002–2007 period is included in the sample.
40
TABLE 2
SFAS 142 Goodwill Impairments by Industry
Industry Name
SIC
Firms Reporting Goodwill
Impairments (a)
% of Total
Sample (b)
Compustat Firms
Reporting Goodwill
(c)
% of Total Compustat Firms with Goodwill
(d)
Mean of Reported
Impairment / Market valuet-1
(e)
Building Construction 15 11 1.8% 116 0.2% 3% Food and Kindred Products 20 12 2.0% 1,003 2.0% 7% Chemicals and Allied Products 28 32 5.2% 3,213 6.5% 11% Rubber and Misc Plastics 30 10 1.6% 374 0.8% 3% Primary Metal 33 13 2.1% 522 1.1% 7% Ind./Comm. Machinery and Computers 35 38 6.2% 2,608 5.3% 7% Electronic Equipment 36 62 10.1% 3,218 6.5% 16% Transportation Equipment 37 14 2.3% 880 1.8% 25% Instruments and Related Products 38 36 5.9% 3,202 6.5% 7% Communication 48 28 4.6% 1,573 3.2% 19% Wholesale Trade - Durable Goods 50 10 1.6% 1,299 2.6% 11% Wholesale Trade - Non-durable Goods 51 11 1.8% 528 1.1% 14% Eating and Drinking Places 58 13 2.1% 791 1.6% 4% Miscellaneous Retail 59 12 2.0% 1,130 2.3% 14% Depository Institutions 60 28 4.6% 5,767 11.7% 2% Business Services 73 82 13.4% 6,895 13.9% 24% Health Services 80 14 2.3% 1,068 2.2% 21% Engineering and Related Services 87 16 2.6% 1,088 2.2% 21% All Other Industries (aggregated) 170 27.8% 14,299 28.9% 7%
612 100.0% 49,574 100.0%
This table presents the industry composition during the sample period 2002-2007, with industries representing 1.5% or more (less than 1.5%) of the sample goodwill impairment firms presented individually (aggregated). We provide the industry composition for both the sample firms reporting goodwill impairments under SFAS 142, as well as all Compustat firms reporting positive goodwill but no goodwill impairment. The final column presents the mean reported goodwill impairment divided by lagged market value for those firms reporting goodwill impairments.
41
TABLE 3
Descriptive Statistics
Observations with
Goodwill Impairment (“GW Impair”)
(N = 612)
Observations with Goodwill but
No Impairment (“All Other”) (N = 49,574)
Tests of Means (GW Impair – All Other)
Test of Medians (GW Impair – All Other)
Variable Mean Median Mean Median Mean p-value Median p-value
SIZE 6,706.960 578.614 9,533.970 583.486 –2,827.010 0.055 –4.870 0.109
FOLLOW 5.724 4.000 5.428 3.000 0.296 0.262 1.000 0.003
BTM 0.753 0.617 0.490 0.445 0.263 0.000 0.172 0.000
ROA –0.095 –0.023 0.002 0.008 –0.097 0.000 –0.031 0.000
GW 0.146 0.101 0.142 0.083 0.004 0.518 0.018 0.912
LAG_GW 0.190 0.144 0.144 0.086 0.046 0.000 0.058 0.000
RESTRUCT_D 0.327 0.000 0.145 0.000 0.182 0.000 0.000 0.000
OTH_IMPAIR_D 0.317 0.000 0.051 0.000 0.266 0.000 0.000 0.000
GW_IMPAIR 0.119 0.023 This table presents descriptive statistics and comparisons across firms that report a goodwill impairment (designated as “GW Impair”) versus all other Compustat firms reporting positive goodwill at the end of the prior quarter but not reporting a goodwill impairment over our sample period (designated as “All Other”). Across all variables for firms reporting goodwill impairments, quarter t refers to the impairment quarter. All p-values are based on two-tailed t-tests. The variables are: SIZE = total assets (in $ millions) at the end of quarter t; FOLLOW = the number of financial analysts (per IBES) following the firm in quarter t; BTM = the book to market ratio at the end of quarter t, with book reflecting any reported goodwill impairment; ROA = income before extraordinary items divided by total assets at the end of quarter t, with both income and total assets
reflecting any reported goodwill impairment; GW = the ratio of goodwill at the end of quarter t to total assets at the end of quarter t, with both goodwill and total assets
reflecting any reported impairment charges; LAG_GW = the ratio of goodwill at the end of quarter t-1 to total assets at the end of quarter t-1;
42
RESTRUCT_D = an indicator variable equal to one for firms reporting restructuring charges in quarter t, and zero otherwise; OTH_IMPAIR_D = an indicator variable equal to one for firms reporting non-goodwill impairments in quarter t, and zero otherwise; and GW_IMPAIR = reported goodwill impairment for quarter t divided by market value of equity at the end of quarter t-1.
43
TABLE 4
Occurrence of Net Insider Selling Over the Two Years Preceding the Goodwill Impairment: Univariate Analyses
Window (month 0 is reporting of
goodwill impairment)
Goodwill Impairment Firms
Non-impairment Firms
Difference in Means (Goodwill Impairment –
Non-impairment)
Difference in Medians (Goodwill Impairment –
Non-impairment)
N Mean Median N Mean Median Mean t-statistic Median z-statistic
(–23, –18) 607 0.435 0.000 4,920 0.364 0.000 0.071 3.410 *** 0.000 3.403 ***
(–17, –12) 626 0.468 0.000 5,216 0.364 0.000 0.104 5.080 *** 0.000 5.067 ***
(–11, –6) 651 0.429 0.000 5,459 0.371 0.000 0.058 2.840 *** 0.000 2.839 ***
( –5, 0) 669 0.371 0.000 5,680 0.367 0.000 0.004 0.180 0.000 0.813 This table presents results from univariate analyses examining the occurrence of net insider selling during the two years preceding the reported goodwill impairment under SFAS 142 across two groups of firms: those reporting goodwill impairments (“Goodwill Impairment Firms”), and not reporting goodwill impairments (“Non-Impairment Firms”). Insider trading is assessed as an indicator variable equal to one for firms that experience net insider trading (i.e., total sales greater than total purchases, calculated using open-market transactions by firm officers) over the indicated six-month trading window, and zero otherwise. Results are presented over four non-overlapping windows: months –23 to –18 (inclusive); months –17 to –12 (inclusive); months –11 to –6 (inclusive); and months –5 to 0 (inclusive). For firms reporting goodwill impairments, the windows are assessed relative to the end of the quarter in which the goodwill impairment is reported, with the month of the impairment designated as month 0. For firms not reporting goodwill impairments, we randomly select a single quarter within our sample reporting period to determine the reporting window. The differences in mean values are based on two-tailed t-tests; the differences in medians are based on two-tailed Wilcoxon tests.
44
TABLE 5
Occurrence of Net Insider Selling Over the Two Years Preceding the Goodwill Impairment:
Multivariate Analyses
Variable
Predicted Sign
Window (–23, –18)
Window (–17, –12)
Window (–11, –6)
Window (–5, 0)
Intercept ? –1.583 *** –1.705 *** –1.954 *** –2.100 *** (–12.870) (–17.917) (–9.423) (–9.816)
GWI_D + 0.121* 0.327 *** 0.187 ** 0.012 (1.351) (3.115) (2.177) (0.117)
SIZE + 0.262 *** 0.276 *** 0.293 *** 0.274 *** (23.840) (25.402) (13.274) (12.684)
BTM – –0.962 *** –0.878 *** –0.591 *** –0.239 *** (–9.617) (–7.822) (–6.030) (–4.088)
RETBH + 0.538 *** 0.486 *** 0.650 *** 0.673 *** (6.080) (6.113) (8.145) (13.343)
RESTRUCT_D + / – 0.166 0.199 ** –0.047 0.151 (1.420) (1.966) (–0.509) (1.223)
OTH_IMPAIR_D + / – –0.119 –0.243 ** –0.088 –0.386 *** (–0.987) (–2.156) (–0.941) (–3.996)
UE_Q0 – –0.025 –0.124 –0.278 *** –0.018 (–1.236) (–1.056) (–3.097) (–0.526)
UE_Q-1 – –0.017 –0.181 *** –0.137 *** 0.016 (–0.348) (–3.869) (–2.684 ) (0.250)
UE_Q-2 – –0.107** –0.044 –0.180 (–2.457) (–0.349) (–1.058)
UE_Q-3 – –0.340* –0.091 –0.335 * (–1.303) (–0.704) (–1.580)
UE_Q-4 – –0.050 –0.155 (–0.333) (–0.929)
UE_Q-5 – –0.234* –0.130 (–1.476) (–1.063)
UE_Q-6 – –0.314 ** (–1.948)
UE_Q-7 – –0.123** (–2.107)
Pseudo R2 0.180 0.180 0.166 0.124
Observations 5,527 5,842 6,110 6,349
This table presents results from multivariate logistic analyses examining the determinants of the occurrence of insider trading during the two years preceding the reported goodwill impairment under
45
SFAS 142. Two groups of firms are included in the sample: those reporting goodwill impairments (“Goodwill Impairment Firms”), and those not reporting goodwill impairments (“Non-Impairment Firms”). The dependent variable is NETSELL_D, an indicator variable equal to one for firms that experience net insider trading (i.e., total sales greater than total purchases, calculated using open-market transactions by firm officers) over the indicated six-month trading window, and zero otherwise. Results are presented over four non-overlapping windows: months –23 to –18 (inclusive); months –17 to –12 (inclusive); months –11 to –6 (inclusive); and months –5 to 0 (inclusive). For firms reporting goodwill impairments (the reporting of which is designated as month 0), the windows are assessed relative to the end of the quarter in which the goodwill impairment is reported. For firms not reporting goodwill impairments, we randomly select a single quarter within our sample reporting period to determine the reporting window. z-statistics are presented in parentheses. ***, ** and * denote statistical significance at 1%, 5% and 10%, respectively, based on one-tail (two-tail) tests for variables having (lacking) directional predictions. Standard errors are cluster-adjusted by time following Petersen (2009). The variables are: NETSELL_D = an indicator variable equal to one for firms that experience net insider selling
(i.e., total sales greater than total purchases) over the indicated six-month trading window, and zero otherwise;
GWI_D = an indicator variable equal to one for firms reporting a goodwill impairment in quarter t, and zero otherwise;
SIZE = log of total assets measured at the end of the indicated six-month window; BTM = the book-to-market ratio measured at the end of the indicated six-month
window; RETBH = the buy-and-hold return measured over the six-month window prior to the
indicated insider trading measurement window; RESTRUCT_D = an indicator variable equal to one for firms reporting restructuring charges in
quarter t (the goodwill impairment quarter), and zero otherwise; OTH_IMPAIR_D = an indicator variable equal to one for firms reporting any non-goodwill
impairments in quarter t (the goodwill impairment quarter), and zero otherwise; and
UE_Qt-k = the unexpected earnings for quarter t-k, where the unexpected earnings are measured as earnings at the end of quarter t-k minus earnings at the end of quarter t-k-4, scaled by market value of equity at the end of quarter t-k-1.
46
TABLE 6
Occurrence of Net Insider Selling Over the Two Years Preceding the Goodwill Impairment:
Subsamples by Quality of Information Environment
Panel A: Low Quality Information Environment Panel B: High Quality Information Environment
Variable
Pred Sign
Window (–23, –18)
Window (–17, –12)
Window (–11, –6)
Window (–5, 0)
Window (–23, –18)
Window (–17, –12)
Window (–11, –6)
Window (–5, 0)
Intercept ? –2.028 *** –2.088 *** –2.213 *** –2.184 *** 0.227 –0.101 0.040 –0.006 (–15.117) (–17.737) (–9.462) (–9.310) (0.901) (–0.627) (0.184) (–0.032)
GWI_D + 0.396 *** 0.729 *** 0.440 *** 0.511 *** 0.191* 0.321 ** 0.260 * –0.091 (3.496) (6.962) (3.040) (2.948) (1.322) (2.288) (1.498) (–0.770)
SIZE + 0.193 *** 0.182 *** 0.206 *** 0.166 *** 0.115 *** 0.143 *** 0.134 *** 0.119 *** (8.241) (9.134) (6.578) (5.008) (3.806) (6.353) (4.558) (4.772)
BTM – –0.510 *** –0.444 *** –0.255 *** –0.101 *** –1.698 *** –1.407 *** –1.474 *** –1.199 *** (–7.140) (–13.133) (–5.141) (–5.341) (–8.108) (–7.075) (–12.831) (–4.154)
RETBH + 0.668 *** 0.495 *** 0.565 *** 0.578 *** 0.361 ** 0.444 *** 0.795 *** 0.640 *** (9.643) (5.477) (5.077) (8.774) (2.059) (2.692) (8.746) (3.350)
RESTRUCT_D + / – –0.018 0.048 –0.189 0.019 0.074 0.109 –0.075 0.139 (–0.118) (0.279) (–0.955) (0.110) (0.501) (1.155) (–0.683) (0.902)
OTH_IMPAIR_D + / – –0.046 –0.174 –0.101 –0.365 ** –0.220 * –0.283 ** –0.095 –0.356 *** (–0.281) (–0.936) (–0.620) (–2.208) (–1.804) (–1.909) (–0.499) (–2.411)
UE_Q0 – –0.009 –0.218 ** –0.178 *** 0.029 –0.107 0.028 –0.187 –0.090 (–0.363) (–2.165) (–3.448) (0.907) (–0.473) (0.330) (–0.237) (–0.572)
UE_Q-1 – 0.062 –0.050 ** –0.068 *** 0.011 –0.795 ** –0.315 ** –0.461 2.161 (1.196) (–2.165) (–3.059) (0.158) (–2.037) (–1.810) (–0.562) (1.969)
UE_Q-2 – –0.103 ** 0.020 –0.115 –0.080 –0.394 * –0.769 (–1.945) (0.297) (–0.950) (–0.290) (–1.414) (–0.654)
UE_Q-3 – –0.349 ** –0.055 –0.308 *** 0.272 –0.126 1.432 (–2.176) (–0.869) (–2.349) (0.686) (–0.319) (2.873)
UE_Q-4 – –0.059 –0.246 ** 0.266 –0.665 ** (–0.958) (–1.690) (0.675) (–2.071)
UE_Q-5 – –0.040 –0.148 * –2.156 *** 2.213 (–0.471) (–1.404) (–5.335) (1.866)
UE_Q-6 – –0.417 *** 1.443 (–2.893) (2.725)
UE_Q-7 – –0.074 * 0.806
47
(–1.516) (2.324)
Pseudo R2 0.108 0.097 0.081 0.053 0.110 0.109 0.112 0.084
Observations 2,991 3,013 3,418 3,607 2,536 2,829 2,692 2,742
This table presents results from multivariate logistic analyses examining the determinants of the occurrence of insider trading during the two years preceding the reported goodwill impairment under SFAS 142. The sample includes two groups of firms: those reporting goodwill impairments, and those not reporting goodwill impairments. Panel A presents results using the sub-sample of firms having a low quality information environment, defined as firms having below median analyst following. Panel B presents results using the sub-sample of firms having a high quality information environment, defined as firms having above median analyst following. Across both panels, the dependent variable is NETSELL_D, an indicator variable equal to one for firms that experience net insider trading (i.e., total sales greater than total purchases, calculated using open-market transactions by firm officers) over the indicated six-month trading window, and zero otherwise. Results are presented over four non-overlapping windows: months –23 to –18 (inclusive); months –17 to –12 (inclusive); months –11 to –6 (inclusive); and months –5 to 0 (inclusive). For firms reporting goodwill impairments, the windows are assessed relative to the end of the quarter in which the goodwill impairment is reported, with the month of the impairment designated as month 0. For firms not reporting goodwill impairments, we randomly select a single quarter within our sample reporting period to determine the reporting window. z-statistics are presented in parentheses. ***, ** and * denote statistical significance at 1%, 5% and 10%, respectively, based on one-tail (two-tail) tests for variables having (lacking) directional predictions. Standard errors are cluster-adjusted by time following Petersen (2009). The variables are: NETSELL_D = an indicator variable equal to one for firms that experience net insider selling (i.e., total sales greater than total
purchases) over the indicated six-month trading window, and zero otherwise; GWI_D = an indicator variable equal to one for firms reporting a goodwill impairment in quarter t, and zero otherwise; SIZE = log of total assets measured at the end of the indicated six-month window; BTM = the book-to-market ratio measured at the end of the indicated six-month window; RETBH = the buy-and-hold return measured over the six-month window prior to the indicated insider trading window; RESTRUCT_D = an indicator variable equal to one for firms reporting restructuring charges in quarter t (the goodwill impairment
quarter), and zero otherwise; OTH_IMPAIR_D = an indicator variable equal to one for firms reporting any non-goodwill impairments in quarter t (the goodwill
impairment quarter), and zero otherwise; and UE_Qt-k = the unexpected earnings for quarter t-k, where the unexpected earnings are measured as earnings at the end of quarter t-
k minus earnings at the end of quarter t-k-4, scaled by market value of equity at the end of quarter t-k-1.
48
TABLE 7
Abnormal Returns to Insider Selling Over the Two Years Preceding the Goodwill Impairment: Univariate Analyses
Insider Trading Window
Abnormal Buy and
Hold Return Window
Buy-and-Hold Return: Net Insider Selling Impairment Firms (NETSELL_D = 1)
Buy-and-Hold Return: All Other
Impairment Firms (NETSELL_D = 0)
Difference in Means (Net Insider Selling –
All Other)
Difference in Medians (Net Insider Selling –
All Other)
N Mean Median N Mean Median Mean t-statistic Median z-statistic
( –8, –3) ( –2, 0) 246 –0.084 –0.059 366 –0.080 –0.082 –0.004 0.190 0.023 0.373
(–11, –6) ( –5, 0) 265 –0.138 –0.123 347 –0.120 –0.139 –0.018 0.720 0.016 0.004
(–14, –9) ( –8, 0) 284 –0.168 –0.162 328 –0.158 –0.181 –0.010 0.330 0.019 0.379
(–17, –12) (–11, 0) 287 –0.204 –0.217 325 –0.134 –0.192 –0.070 2.020 ** –0.025 –1.244
(–20, –15) (–14, 0) 277 –0.223 –0.234 335 –0.115 –0.209 –0.108 2.740 *** –0.025 –2.187 **
(–23, –18) (–17, 0) 268 –0.234 –0.277 344 –0.105 –0.197 –0.129 3.010 *** –0.080 –2.246 **
This table presents univariate comparisons of abnormal buy-and-hold returns (measured as the firm raw return less the value-weighted market
return) to insider trades for two groups of firms reporting goodwill impairments under SFAS 142. The first group is firms, in which insiders are
net sellers (i.e., the variable NETSELL_D = 1), and is designated as “Net Insider Selling Impairment Firms.” The second is firms, in which
insiders are not net sellers (i.e., the variable NETSELL_D = 0), and is designated as “All Other Impairment Firms.” Insider trades (defined as
open-market transactions by firm officers) are measured over six-month rolling windows covering the two years prior to the goodwill impairment
event. The buy-and-hold abnormal returns to insider trades are computed over windows starting directly after the end of the insider trading
window and ending with the last month of the goodwill impairment quarter (i.e., month 0). ***, ** and * denote statistical significance at 1%, 5%
and 10%, respectively, based on two-tailed t-tests of means or two-tailed z-tests of medians.
TABLE 8
Abnormal Returns to Insider Selling Over the Two Years
Preceding the Goodwill Impairment: Multivariate Analyses
Variable
Pred Sign
Return Window (–17, 0)
Return Window (–14, 0)
Return Window (–11, 0)
Return Window (–8, 0)
Return Window (–5, 0)
Return Window (–2, 0)
Intercept ? –0.324*** –0.253** –0.277*** –0.321*** –0.313*** –0.179*** (–2.740) (–2.500) (–3.030) (–3.340) (–4.560) (–4.660)
NETSELL_D – –0.125** –0.090*** –0.073** –0.029 –0.027 –0.020 (–2.270) (–2.380) (–2.290) (–0.780) (–0.780) (–0.870)
SIZE + / – 0.029** 0.018 0.024** 0.026** 0.024*** 0.015*** (2.070) (1.450) (2.200) (2.630) (2.780) (3.000)
BTM + / – 0.102 0.097* 0.053 0.054* 0.099** 0.041** (1.610) (1.760) (1.070) (0.970) (2.580) (2.200)
RETBH + / – 0.147** 0.141* 0.149** 0.158** 0.186*** 0.184*** (2.260) (1.710) (2.520) (2.330) (3.100) (2.920)
RESTRUCT_D + / – –0.006 –0.035 –0.054 –0.044 –0.040 –0.006 (–0.130) (–0.930) (–1.500) (–1.680) (–1.600) (–0.300)
OTH_IMPAIR_D + / – –0.118*** –0.119*** –0.080** –0.072** –0.047 –0.031 (–3.000) (–3.040) (–2.580) (–2.020) (–1.210) (–1.030)
UE_Q0 + –0.168 –0.164 –0.107 –0.144 –0.096 –0.028 (–0.990) (–1.060) (–1.220) (–1.510) (–1.510) (–0.530)
UE_Q-1 + 0.445** 0.290** 0.380*** 0.132 0.049 (1.910) (1.790) (2.390) (0.890) (0.340)
UE_Q-2 + 0.058 0.0 71 –0.051 –0.028 (1.050) (0.860) (–0.680) (–1.080)
UE_Q-3 + 0.622** 0.621*** 0.854*** (1.910) (3.700) (4.310)
UE_Q-4 + 0.227* 0.210* (1.430) (1.360)
UE_Q-5 + 0.645* (1.310)
Pseudo R2 0.083 0.075 0.082 0.059 0.064 0.059 This table presents results from multivariate analyses examining abnormal buy-and-hold returns to insider trades for two groups of firms reporting goodwill impairments under SFAS 142. The first group is firms, in which insiders are net sellers (i.e., the variable NETSELL_D = 1). The second is firms, in which insiders are not net sellers (i.e., the variable NETSELL_D = 0). Insider trades (defined as open-market transactions by firm officers) are measured over six-month rolling windows covering the two years prior to the goodwill impairment event. The dependent variable is ABNORMAL_RET, the buy-and-hold abnormal returns to insider trades computed over windows starting directly after the end of the insider trading window and ending with the last month of the goodwill impairment quarter (i.e., month 0). ***,
50
** and * denote statistical significance at 1%, 5% and 10%, based on one-tail (two-tail) tests for variables having (lacking) directional predictions. Standard errors are cluster-adjusted by time following Petersen (2009). For all regressions, N = 612. The variables are: ABNORMAL_RET = the buy-and-hold abnormal returns (measured as the firm raw return less the
value-weighted market return) to insider trades, computed over windows starting directly after the end of the insider trading window and ending with the last month of the goodwill impairment quarter (month 0);
NETSELL_D = an indicator variable equal to one for firms that experience net insider selling (i.e., total sales greater than total purchases) over the indicated six-month trading window, and zero otherwise;
SIZE = log of total assets measured at the end of the indicated six-month window; BTM = the book-to-market ratio measured at the end of the indicated six-month
window; RETBH = the buy-and-hold return measured over the six-month window prior to the
indicated insider trading measurement window; RESTRUCT_D = an indicator variable equal to one for firms reporting restructuring charges in
quarter t (the goodwill impairment quarter), and zero otherwise; OTH_IMPAIR_D = an indicator variable equal to one for firms reporting any non-goodwill
impairments in quarter t (the goodwill impairment quarter), and zero otherwise; and
UE_Qt-k = the unexpected earnings for quarter t-k, where the unexpected earnings are measured as earnings at the end of quarter t-k minus earnings at the end of quarter t-k-4, scaled by market value of equity at the end of quarter t-k-1.
TABLE 9
Abnormal Returns to Insider Selling Over the Two Years Preceding the Goodwill Impairment:
Subsamples by Quality of Information Environment
Panel A: Low Quality Information Environment Panel B: High Quality Information Environment
Variable (Predict Sign)
Return Window (–17, 0)
Return Window (–14, 0)
Return Window (–11, 0)
Return Window (–8, 0)
Return Window (–5, 0)
Return Window (–2, 0)
Return Window (–17, 0)
Return Window (–14, 0)
Return Window (–11, 0)
Return Window (–8, 0)
Return Window (–5, 0)
Return Window (–2, 0)
Intercept –0.273 * –0.197 –0.254** –0.331 *** –0.324 *** –0.222 *** –0.576*** –0.540*** –0.657* ** –0.565 *** –0.496 *** –0.201 ** (?) (–1.90) (1.45) (–2.51) (–2.85) (–4.58) (–4.90) (–3.58) (–4.18) (–5.82) (–5.59) (–4.23) (–1.85)
NETSELL_D –0.174 ** –0.106 * –0.095 * –0.042 –0.060* –0.029 –0.026 –0.065 * –0.012 –0.005 0.019 0.017 (–) (–2.08) (–1.68) (–1.79) (–0.63) (–1.44) (–0.79) (–0.52) (–1.50) (–0.29) (–0.14) (0.42) (0.46)
SIZE 0.025 0.015 0.032 ** 0.033 *** 0.031 *** 0.021 *** 0.051** 0.050 *** 0.061 *** 0.054 *** 0.042 *** 0.016 (+/–) (1.37) (0.92) (2.61) (3.25) (3.79) (3.90) (2.71) (3.85) (4.93) (4.30) (2.95) (1.45)
BTM 0.105 0.086 0.035 0.066 0.107 *** 0.045 0.151 0.114 0.096* 0.027 0.112** 0.056* (+/–) (1.48) (1.46) (0.66) (0.99) (2.97) (1.65) (1.56) (1.23) (1.68) (0.46) (2.04) (1.04)
RETBH 0.114 0.129 0.115* 0.127 0.253 *** 0.208 *** 0.194 ** 0.138 0.078 0.096 0.048 0.125 (+/-) (1.16) (1.18) (1.72) (1.42) (3.02) (3.50) (2.52) (1.63) (0.78) (1.27) (0.67) (1.23)
RESTRUCT_D 0.016 -0.027 –0.036 –0.037 0.001 0.062 –0.014 –0.012 –0.034 –0.017 –0.052 * –0.068*** (+ / –) (0. 23) (-0.42) (–0.47) (–0.72) (0.02) (1.62) (–0.39) (–0.25) (–1.07) (–0.76) (–1.81) (–2.78)
OTH_IMPAIR_D –0.112 * –0.129 * –0.062 –0.050 –0.041 –0.015 –0.085 ** –0.092 ** –0.096** –0.102 ** –0.053 –0.053 (+ / –) (–1.71) (–1.82) (–1.18) (–0.91) (–0.87) (–0.39) (–2.19) (–2.06) (–2.28) (–2.31) (–1.34) (–1.36)
UE_Q0 –0.107 –0.101 –0.067 –0.098 –0.042 –0. 001 –0.540 –0.522 –0.415 –0.313 –0.345 –0.217 (+) (–0.63) (–1.11) (–0.83) (–1.06) (–0.88) (–0.17) (–4.19) (–4.41) (–5.00) (–5.12) (–4.47) (–2.70)
UE_Q-1 0.206 0.069 0.186 ** –0.024 –0.105 1.974*** 1.675*** 1.521*** 1.082 *** 1.512*** (+) (1.09) (0.67) (1.72) (–0.20) (–1.19) (6.93) (3.98) (2.71) (3.01) (5.43)
UE_Q-2 0.072** 0.083 –0.027 –0.018 2.834 *** 1.776 0.731 1.170* (+) (1.71) (1.24) (–0.53) (–0.60) (2.65) (1.22) (0.67) (1.40)
UE_Q-3 0.451* 0.482*** 0.6028*** 2.264*** 2.731*** 2.986*** (+) (1.50) (3.05) (4.03) (3.23) (2.49) (3.84)
UE_Q-4 0.195 0.199 –0.621 –0.941 (+) (1.05) (1.19) (–0.85) (–1.25)
UE_Q-5 0.594 0.144 (+) (1.12) (0.10)
Pseudo R2 0.089 0.077 0.073 0.061 0.109 0.096 0.268 0.258 0.230 0.226 0.175 0.104
N 340 336 327 333 324 332 272 276 285 279 288 280
52
This table presents results from multivariate analyses examining abnormal buy-and-hold returns to insider trades for two groups of firms reporting goodwill impairments under SFAS 142. The first group is firms, in which insiders are net sellers (i.e., the variable NETSELL_D = 1). The second is firms, in which insiders are not net sellers (i.e., the variable NETSELL_D = 0). Insider trades (defined as open-market transactions by firm officers) are measured over six-month rolling windows covering the two years prior to the goodwill impairment event. Panel A presents results using the sub-sample of firms having a low quality information environment, defined as firms having below median analyst following. Panel B presents results using the sub-sample of firms having a high quality information environment, defined as firms having above median analyst following. Across both panels, the dependent variable is ABNORMAL_RET, the buy-and-hold abnormal returns to insider trades computed over windows starting directly after the end of the insider trading window and ending with the last month of the goodwill impairment quarter (i.e., month 0). ***, ** and * denote statistical significance at 1%, 5% and 10%, based on one-tail (two-tail) tests for the cases where we have (do not have) directional predictions. Standard errors are cluster-adjusted by time following Petersen (2009). The variables are: ABNORMAL_RET = the buy-and-hold abnormal returns to insider trades, computed over windows starting directly after the end of the
insider trading window and ending with the last month of the goodwill impairment quarter (month 0); NETSELL_D = an indicator variable equal to one for firms that experience net insider selling (i.e., total sales greater than total
purchases) over the indicated six-month trading window, and zero otherwise; SIZE = log of total assets measured at the end of the indicated six-month window; BTM = the book-to-market ratio measured at the end of the indicated six-month window; RETBH = the buy-and-hold return measured over the six-month window prior to the indicated insider trading measurement
window; RESTRUCT_D = an indicator variable equal to one for firms reporting restructuring charges in quarter t (the goodwill impairment
quarter), and zero otherwise; OTH_IMPAIR_D = an indicator variable equal to one for firms reporting any non-goodwill impairments in quarter t (the goodwill
impairment quarter), and zero otherwise; and UE_Qt-k = the unexpected earnings for quarter t-k, where the unexpected earnings are measured as earnings at the end of quarter t-
k minus earnings at the end of quarter t-k-4, scaled by market value of equity at the end of quarter t-k-1.